Fluoroscopy as an imaging means for computer-assisted surgical navigation

Biomedical Paper

Fluoroscopy as an Imaging Means for Computer-Assisted Surgical Navigation

R. Hofstetter, Dipl.-Ing., M. Slomczykowski, M.D., Ph.D., M. Sati, Ph.D.,and L.-P. Nolte, Ph.D.-Ing.

Maurice E. Muller Institute for Biomechanics, University of Bern, Bern, Switzerland

ABSTRACT Objective: Intraoperative fluoroscopy is a valuable tool for visualizing underlyingbone and surgical tool positions in orthopedic procedures. Disadvantages of this technology includethe need for continued radiation exposure for visual control, and cumbersome means of alignment.The purpose of this article was to highlight a new concept for a computer-assisted freehandnavigation system that uses single intraoperatively acquired fluoroscopic images as a basis forreal-time navigation of surgical tools.

Materials and Methods: Optoelectronic markers are placed on surgical tools, a patientreference, and the fluoroscope to track their position in space. Projection properties of the fluoro-scope are acquired through an initial precalibration procedure using a tracked radiopaque phantomgrid. Corrections are applied to compensate for both the fluoroscope’s image intensifier distortionsand the mechanical bending of the C-arm frame. This enables real-time simulation of surgical toolpositions simultaneously in several single-shot fluoroscopic images. In addition, through optoelec-tronically tracked digitization of a target viewpoint, the fluoroscope can be numerically aligned atprecise angles relative to the patient without any X-ray exposure.

Results: This article shows the feasibility of this technology through its use in cadaver trialsto perform the difficult task of distal locking of femoral nails. Comp Aid Surg 4:65–76 (1999). ©1999Wiley-Liss, Inc.

Key words: C-arm, fluoroscopic image registration, computer-integrated surgery, distal locking,femoral nail

INTRODUCTIONMobile fluoroscopic devices (C-arms) are an inte-gral part of the standard equipment used in ortho-pedic surgery to provide real-time feedback of boneand surgical tool positions. A constant or pulsedmode allows control of surgical actions, and singlebuffered images can be used for diagnosis andverification. Although C-arms provide very valu-able in situ image information, their use results inexposure of the patient and operating room staff toradiation. This is especially true when constant and

higher-frequency pulsed modes are used for longerperiods of time while the surgeon is operating closeto the field of view. Since fluoroscopy is a two-dimensional (2-D) imaging technique, it cannotprovide depth information within a single image.This makes the control of complex three-dimen-sional (3-D) manipulations difficult and requiresrepeated use of the fluoroscope at different viewingangles.

Currently available computer-assisted ortho-

Received October 23, 1998; accepted May 3, 1999.

Address correspondence/reprint requests to: R. Hofstetter, Maurice E. Mu¨ller Institute for Biomechanics, University of Bern,Murtenstrasse 35, CH-3010 Bern, Switzerland. E-mail: [email protected].

Computer Aided Surgery 4:65–76 (1999)

©1999 Wiley-Liss, Inc.

pedic surgery systems are generally based on 3-Dimage data sets that are acquired preoperatively ina computed tomography (CT) scanner. Multipleviews of the imaged anatomy are then generatedand brought to the computer screen where the po-sitions of surgical tools are represented.1,2 Thesesystems provide the missing link between the anat-omy and medical images by visualizing the posi-tions of surgical tools relative to the image data,and they have proven to be very useful for certainorthopedic interventions such as pedicle screw in-sertion. However, for other surgeries, e.g., in thefield of osteosynthesis, it is often not possible tojustify acquisition of a CT data set because of bothcost and time constraints.

This is especially true for distal locking of theunreamed femoral nail, one of the most difficultsteps in the process of femoral fracture reduction.During this closed procedure, the C-arm is pres-ently the only aid to guiding a surgical drill throughthe locking holes of the intramedullarly placed nail.In the first step of the conventional approach, anaccurate C-arm alignment has to be achieved toobtain a view through the locking holes, reproduc-ing them as perfect circles. This trial-and-errormethod requires patience and sound experience inC-arm handling. Next, a special radiolucent drill isguided under constant C-arm control through thebone and the locking holes. Note that here thesurgeon is performing a 3-D surgical action usingonly one 2-D view, which often leads to errors.Furthermore, these two steps are responsible for ahigh percentage of the total X-ray exposure duringthis procedure.

This has led us to integrate the C-arm into acomputer-assisted surgery system and to providesurgical navigation based on intraoperatively ac-quired single-shot images. Two different featuresrequired for the described locking procedure havebeen included: navigation of surgical tools in C-arm images, and navigation of the C-arm to obtainoptimal views of the anatomy.

The concept required the development of anaccurate method for automatic registration of fluo-roscopic images which did not interfere with rou-tine operation procedures. This article describes thegeneral methods of this technology and demon-strates its feasibility for use in locking femoralnails.

Previous WorkThe first attempts to use fluoroscopy in a system forimage-guided surgery were made in 1989, when thetechnique was applied to kidney stone removal.3 A

passive robotic manipulator was used to measurethe positions of surgical tools and a phantom forC-arm image registration. For a project involvingtotal hip replacement,4 the use of registered C-armimages has been proposed for coregistering CT-image data or 3-D models to a surgical robot.

Hamadeh et al.5,6 and Weese et al.7 usedseveral images of a calibrated C-arm for coregis-tration of spinal CT data. Brack et al.8 used acalibrated C-arm to guide saw cuts for knee arthro-plasty surgery; Brandt et al.9 proposed the use ofintraoperatively acquired and registered C-arm im-ages to guide active or semiactive surgical robots inpositioning orthopedic implants. Viant et al.10 re-cently published a method based on biplane-regis-tered C-arm images for determining the trajectoriesof distal locking screws through a femoral nail. Inall of these papers, C-arm calibration is performedseparately for each intraoperative image by placinga calibration phantom in the field of view. We havechosen an approach involving C-arm precalibrationthat avoids the use of calibration phantoms in everyimage and is thus advantageous for routine clinicaluse. A first prototype using our concept was brieflydescribed in 1997.11

Because the C-arm’s X-ray projection can bemodeled as an optical camera system (extrinsiccalibration), our work has been inspired by thecalibration of video cameras, especially in indus-trial and robotic applications where much work hasbeen done.12 Martins introduced an extrinsic bi-plane calibration method using coplanar-placedcalibration markers and studied its accuracy.13 Adetailed mathematical approach for calibrating avideo camera using a pinhole camera model wasgiven by Gembran et al.,14 and a more complicatedmethod using B-spline instead of linear approxima-tion was proposed by Champleboux and col-leagues.15 Accurate calibration of the C-arm re-quires consideration of distortions caused by theimage intensifier process that can be separatelymodeled (intrinsic calibration). In earlier publishedreports,16–18 pincushion distortions were handledthrough a few parameters of a physical model andmethods for correction were proposed. Mansbach19

developed a method to calibrate an optical systemwithout distinguishing between intrinsic and extrin-sic distortions.

MATERIALS AND METHODSAn optoelectronic position sensor (Optotrak 3020;Northern Digital, Waterloo, Ontario, Canada) isused to track the position of surgical tools, a patientreference, and the image intensifier of the C-arm

66 Hofstetter et al.: Fluoroscopy for Surgical Navigation

within the region of the operating table. Each com-ponent is equipped with infrared (IR) light-emittingdiode (LED) markers which define local coordinatesystems (COS) (Fig. 1). The position sensor sup-plies data needed to perform coordinate transfor-mations between these local coordinate systems.This data is passed on to a workstation computer(Ultra 1; Sun Microsystems, Mountain View, CA).An off-the-shelf video framegrabber board (SLIC-Video; Osprey Systems, Cary, NC) allows loadingof gray-scale images from the video buffer of theC-arm (BV22; Philips, Hamburg, Germany) to theworkstation.

Our objective was to arrive at a system thatworks as follows: The surgeon acquires a numberof single C-arm views from different orientationsthat are all displayed on the workstation screen. Acomputer-generated projection of a surgical tool isthen displayed on each image. This is equivalent toits representation under conventional constant flu-oroscopic control. An update rate of 10 Hz enablesreal-time navigation in up to four C-arm imagessimultaneously.

The underlying computations are based on achain of transformations as given in Equation 1. Letthe tip of a drill bit be described by a 3-D vectorvT.This vector is constant in its local tool coordinatesystem (T-COS; note that the subscripts indicatethe COS in which the vectors are reported). Con-sidering the position of the surgical object, whichcan either be the patient or an implant,vT is firsttransformed tovP in the patient/implant coordinate

system (P-COS). The transformation matrixTT,P,transforming coordinates from the T-COS to theP-COS, is provided in real time by the positionsensor; then,vP is transformed into the C-arm im-age intensifier coordinate system (A-COS). Thetransformation matrixTP,A describes the positionof the C-arm image intensifier relative to the patientat the time of image acquisition. The position of theC-arm must be acquired directly after its triggerbutton has been pressed. Note that every C-armimage loaded onto the computer requires a separateTP,A. Finally vA is transformed tovI located in the2-D, pixel-based C-arm image coordinate system(I-COS).

vTf vPf vA f vI

with

vP 5 vT z TT,P, vA 5 vP z TP,A, vI 5 vA z TA,I (1)

The associated transformationTA,I that mod-els the actual X-ray projection is based on a conebeam projection model as shown in Figure 2. It wasused by Mansbach19 to calibrate video cameras forrobotic applications. Let the center of the X-raysource be represented by a 3-D vector,fA (focalpoint). Two vectors,rA andcA, that point along therising row and column coordinates of the digitalimage define the orientation of the image planeonto which the C-arm image is projected. The 2-Dvector pI, present in the I-COS, represents thepiercing point of an image plane normal throughfA.

To transform a 3-D coordinatevA into a 2-DcoordinatevI, a unit vectorsA is first calculated

Fig. 2. Linear cone beam projection model as used forextrinsic C-arm calibration. The X-ray is emitted at locationfA and projects a pointvA, representing a tool tip, onto theimage plane asvI. pI is defined by the point where rays passdirectly normal through the plane.

Fig. 1. System setup with components and their associ-ated local coordinate systems (COSs).Top left: C-arm andthe image intensifier.Top middle: Position sensor.Topright: Example with a surgical instrument.Bottom left:Patient representing the surgical object.Bottom right: C-arm image as displayed on the monitor; this is a 2-D COS.

Hofstetter et al.: Fluoroscopy for Surgical Navigation 67

(Equation 2) that points in the direction of theappropriate X-ray beam.

sA 5vA 2 f A

uvA 2 f Au (2)

In a second step,sA is projected onto theimage plane through two scalar products (Equation3). The result is a 2-D coordinate, which is added tothe piercing pointpI.

vIrow 5 sA z rA 1 prow

vIcol 5 sA z cA 1 pcol

pI 5 S prow

pcolD, vI5S vIrow

vIcolD (3)

The vectorsfA, rA, cA, andpI are projectionparameters and describe the properties of the C-arm. While the vectorsrA and cA are assumed toremain constant in the A-COS for the lifetime ofthe C-arm, we have found that the X-ray sourcepositionfA and the related piercing pointpI changeon most C-arm models owing to mechanic defor-mations of the C-arm frame. However, all projec-tion parameters are determined in an initial extrin-sic C-arm calibration procedure, which is repeatedperiodically when required. The method used tocompensate for the C-arm frame deformations isdescribed below.

Extrinsic Calibration

The calibration is based on a plate containing ra-diopaque spherical markers arranged in a rectangu-lar grid. This plate is imaged twice by the C-arm ata proximal and distal position relative to the imageintensifier.

To obtain the 3-D positions of each sphere,the calibration plate is instrumented with LEDmarkers defining a local COS (C-COS). Threereference indents,iC1, iC2, andiC3, locally definethe position of the rectangular grid (Fig. 3). Thecoordinates of these points are digitized withrespect to the C-COS using an optoelectronicallytracked pointer. After these incision points havebeen transformed from the C-COS into the A-COS, the coordinates of all sphere centers aregenerated. From here on, we will refer to the 3-Dsphere centers of the proximal (vAProx) and thedistal (vADist) calibration image in the A-COS asthe original coordinates. The corresponding im-age coordinatesvIProx and vIDist (I-COS) of theprojections of the spheres are obtained throughimage analysis (Fig. 4). By sorting according totheir positions, every image coordinate can beassigned to a corresponding original coordinate.For each calibration image, a 33 3 transforma-tion matrix AProx, ADist is set up. This allowstransformation of a 2-D image coordinatevI into3-D original coordinatesvAProx and vADist, lo-cated in the plane of the calibration plate usingthe following equations:

Fig. 3. Setup for the extrinsic C-arm calibration. A plate containing steel spheres is imaged in a proximal and a distal positionwith respect to the C-arm image intensifier. The 3-D centers of the spheresvAProx, vADist are determined through optoelectronictracking.


vAProx 5 AProx z S vIProxx

vIProxy

1D,

vADist 5 ADist ? S vIDistx

vIDisty

1D (4)

AProx is calculated by setting up an overde-termined system of equations using allvAProx andvIProx of a calibration image. A solution with aleast-squares error is found using the matrixpseudoinverse:

AProx 5 VAProx z V IProxT z (V IProx z V IProx

T )21 (5)

where the matrixVAProx contains all column vec-torsvAProx and the matrixVIProx all column vectorsvIProx with the z-components set to 1. The samemethod is used to calculateADist.

By means of Equation 4, it is then possible tofind the corresponding coordinatesvAProx andvADist for anyvI. The lines formed by joining eachvAProx, vADist represent the X-ray beams belongingto each image coordinatevI (Fig. 3). All X-raybeams ideally intersect at the focal point of thesystem—in this case, the center of the X-ray sourcefA. This point is found by minimizing its distance toall lines of sight corresponding to the spheres of theproximal calibration image using a least-squareserror minimization algorithm.14

The remaining projection parameters,rA, cA

and pI, are determined by setting up a system ofequations based on Equations 2 and 3, with all pairsof image and original coordinates for both calibra-tion images. A matrix pseudoinverse technique isused to solve this overdetermined system of equa-tions. A detailed description of the mathematics canbe found in Gembran et al.14

The errors included during this precalibrationstep have a direct and systematic influence on thenavigation error. Errors are introduced through thetwo optoelectronic-based coordinate transforma-tions, which are required to measure the 3-D cen-ters of the spheres in the two calibration-plate po-sitions. The tracking system provides coordinatedata with a total RMS error of about 0.2 mm perLED marker.20 Roughly estimated for the use ofoptimized LED marker shields on both the imageintensifier and calibration plate, this introduces amean error of 0.4 mm for navigation within thevolume between proximal and distal calibrationplates. Additional errors are caused by the imageanalysis algorithm detecting the centers of thespheres in the calibration images. However, it wasfound through informal testing that, owing to thehigh number of markers, this error is negligible.Overall accuracy analysis is described later in Sys-tem Accuracy Study.

C-Arm Frame DeformationsBecause of the significant weight of the imageintensifier and the X-ray source unit, the C-armframe is subject to variable states of stress accord-ing to the C-arm’s orientation. This may causesignificant frame deformations when the C-arm ismoved. With one of the C-arm models available forour studies, we experimentally measured a maxi-mum movement offA to be 9.8 mm relative to theA-COS. The C-arm frame must therefore be con-

Fig. 5. Position-sensing concept with corresponding co-ordinate systems to compensate for errors related to C-armframe deformations.

Fig. 4. C-arm image of the calibration plane in proximalposition. The 2-D center coordinates of the sphere markersare automatically detected through image analysis andmarked by a white cross.


sidered to act as a nonrigid body with changingprojection parametersfA andpI.

To avoid related inaccuracies, these deforma-tions are measured using the optoelectronic posi-tion sensor. A set of LED markers is attached to theX-ray source chassis of the C-arm (Fig. 5) to definea local coordinate system S-COS. Our strategy is tostore deformations preoperatively over all possibleC-arm orientations in a lookup table (deformationcalibration). During the surgical intervention, whenthe S-COS markers are not visible for the positionsensor, the stored deformation parameters corre-sponding to a given C-arm position are used toupdate the extrinsic projection parametersfA andpI. In a preliminary study with two different C-armmodels, we found that deformations of the framecaused by changes in its position are within theelastic range of the material, and mechanical partssuch as bearings add no significant hysteresis. Wetherefore assume that the C-arm frame deformationis a repeatable function of its orientation relative togravity.

The spatial orientation of the C-arm relativeto gravity can be expressed by two angles,a andb(Fig. 5). An optoelectronic weight-based gravitymeasurement tool has been designed to provide thesystem with a vertical gravity vectorgA as a refer-ence orientation (Fig. 6).a andb can be calculatedfrom the orientation of this gravity vector relativeto the image intensifier.

The corresponding focal point positionfA(a,b) is measured in the following way:

● At the time of extrinsic calibration, the initialvalue offA is transformed from the A-COS to

fS in the S-COS, where it is assumed to remainconstant.

● With the C-arm in a position (a,b), fS istransformed back from the S-COS into theA-COS using a transformation matrix pro-vided by the position sensor.

To generate the lookup table, the C-arm ismoved in 10° intervals to all orientations requiredin an operation, and values ofa, b, andfA(a,b) arestored.

When an image is acquired during an opera-tion, the C-arm orientation anglesas and bs aremeasured. Four entries closest toas andbs are thenidentified in the lookup table. A bilinear interpola-tion between the related four values offA(a,b) isused to compute the focal pointfA(as,bs) valid forthe currently acquired image.

To recalculate the piercing pointpI for thenew value of fA(as,bs), Equations 2 and 3 aresolved forpI with vI 5 (0,0).vA is calculated usingEquation 4 for the proximal calibration image. Weassume thatAProx remains constant for all possiblefA(a,b) because the proximal calibration plate isplaced directly on the image intensifier input phos-phor.

Efforts were made to keep errors originatingfrom this method of C-arm deformation compen-sation at a low level. These errors are mainlycaused by the two involved optoelectronic-basedcoordinate transformations between the A-COSand the S-COS, in conjunction with the large dis-tance of the associated marker shields. We opti-mized the size and position of these marker shieldsand experimentally isolated the related errors. Wetherefore mounted both marker shields at the endsof a nondeforming metal bar at a distance equal totheir positions on the C-arm. Ideally, the trackingsystem should provide a constant position offA forevery orientation of this rigid body in space. Withoptimized marker shields, we obtained a maximumvariation for the position offA of 0.8 mm.21

Image Intensifier Distortions

Video-fluoroscopic images often have distortionsthat originate from the image intensification pro-cess. Two main sources for these distortions havebeen distinguished:

● Stationary pin-cushion distortions arise fromboth the spherical shape of the image intensi-fier input phosphor, onto which the X-rays areprojected, and the optical characteristics of thevideo camera.

Fig. 6. Weight-based optoelectronic gravity measurementtool.


● Distortions of more arbitrary shape that de-pend on the position of the image intensifierrelative to the earth’s magnetic field and otherexternal magnetic fields are caused by theelectron optics of the photo multiplier.17

Our approach is to undistort the image so thatit fits into the linear projection model. The station-ary component of the distortion is removed in anintrinsic precalibration procedure using a local lin-ear distortion correction method.

The centers of the markers in the proximalextrinsic calibration image represent actual localvalues of the distortions. Reference values for anundistorted image are calculated by transformingthe 3-D coordinates of the spheres forming the gridinto the C-arm image (I-COS), using the projectionalgorithm described above (Eqs. 2 and 3) with theparameters obtained during the extrinsic calibra-tion. This guarantees an exact adjustment of thedistortion correction to the extrinsic calibration. Forevery marker, a displacement vector is calculatedthat points from a position in a distorted image to acorresponding position in an undistorted image.

A bilinear local interpolation is performed forevery pixel of the image to identify its displace-ment. The grid of distorted markers is thereforedivided into triangles as shown in Figure 7. Thedisplacement of every pixel within a triangle is

interpolated using displacement values of the mark-ers at the three edges and is then rounded to anearest integer neighbor. For reasons relating tocomputing speed, the displacement of all C-armimage pixels is precalculated and stored in a table,and this is applied to every image obtained duringan operation. Remaining gaps are finally filled us-ing linear gray-value interpolation.

When applying this method, errors are causedby the fact that nonlinear distortions are compen-sated for using linear interpolation. To ascertainthis error experimentally, we imaged a phantomwith 30 steel spheres (2 mm in diameter) arrangedin a straight line, and applied the described distor-tion correction algorithm. Image analysis softwarewas then used to determine the centers of thespheres in the corrected image. An ideal straightline was obtained from these data using a linearregression algorithm. The mean deviation of thepreserved coordinates from this ideal line was 1.2pixels, with a standard deviation of 1.0 pixel for atotal of five trials.

Further errors originate from distortionscaused by external magnetic fields, which cannotbe compensated for by precalibration. They appearonce the C-arm is moved to positions differingfrom that at the time of calibration, however, theC-arm under investigation here was not very sen-sitive to magnetic field distortions. The effect ofthis error on overall system accuracy is reported inSystem Accuracy Study.

C-Arm AlignmentA major difficulty in handling a C-arm during anoperation is ensuring its accurate alignment relativeto the patient’s anatomy to obtain the desired view.In current clinical practice, this is accomplished byrepeatedly acquiring verification images or by us-ing the constant imaging mode. To reduce the re-sultant radiation exposure, we propose a computer-based C-arm alignment method.

The basic concept is to define the desiredC-arm orientation with respect to a patient refer-ence using a pointing device. Defined points arerepresented by a graphical object that is renderedthrough the cone beam projection model onto avirtual C-arm image. This preview mode allowsC-arm alignment without radiation exposure.

For example, two C-arm views must be wellaligned when preparing the holes to lock an un-reamed femoral nail. One view must be orientedexactly through the locking holes so that they ap-pear as ideal circles, while the second view has tobe aligned perpendicular to this hole. In our system,

Fig. 7. Triangle-based displacement interpolation for dis-tortion correction. Correction vector from pixel inside of topleft triangle obtained from linear interpolation of vertexcorrections.


the patient/implant reference is preoperatively fixedto the nail (nail reference). Prior to nail insertion,the positions and orientations of the locking holesare defined with respect to the nail reference usinga special optoelectronic pointing tool. When thenail has been implanted and the C-arm is ready tobe aligned, these landmarks are visualized in real-time before obtaining an image. Each locking holeis represented by a pair of small circles, a longstraight line specifies the position of the nail axis(Fig. 8), and the large circle predicts the field ofview of the C-arm image. A perfect C-arm align-ment is achieved when two circles belonging to ahole are matched and located in the center of thefield of view. Fine-tuning of the C-arm rotationscan be performed through feedback from the twosliders at the bottom of the interface. Once animage aligned directly through the hole is obtained(Fig. 9, right), the system can be switched to guidethe C-arm view perpendicular to the hole axis (Fig.9, left).

User InterfaceA simple graphical user interface was implementedto allow easy use of the system in the operatingtheater. The image acquisition mode enables theuser to load and display up to nine C-arm images.Four of these are displayed in the navigation mode

in which the surgeon can control the position of theactive surgical tool. A zoom on two of these imagescan be seen in Figure 9, showing the pseudo 3-Dnavigation obtained through simultaneous multi-planar views. Note that the drill is represented as acorrespondingly thick line, and that an optional“trajectory line” (dotted line) helps surgeons pre-dict where the drill will pass with the current ori-entation. The system can be easily configured tovisualize any defined landmark (Fig. 8) and can beused to navigate within any rigid bone. It is there-fore a generalized tool, and is rapidly adaptable tonew applications.

System Accuracy StudyTo verify the total system accuracy, i.e., the accu-racy with which a surgical tool is represented onacquired C-arm images, a dedicated optoelectroni-cally tracked phantom was constructed. This accu-racy phantom contains eight spherical steel markers(2 mm in diameter), aligned in two planes (Fig. 10).The coordinates of the sphere centers relative to theoptoelectronic markers were measured with a max-imum error of 0.2 mm. Imaging this phantom al-lows verification of the system accuracy by com-paring the actual sphere representations to thosecalculated by the system. For visual estimation ofthe errors, small crosses are superimposed on theimage in the calculated sphere center position: Ifthe cross is located outside the associated sphere,the error is.1 mm. An exact error quantification,as required for an accuracy study, is performed bymetric measurements on image printouts.

The C-arm used for the accuracy study (Com-et Telam C125, Liebefeld, Switzerland) was cali-brated using a calibration plate with 137 markersand a grid width of 10 mm. We then acquired atotal of 40 images with the C-arm in various posi-tions and at various distances from the described

Fig. 9. Real-time image interactive guidance of surgicaldrill used to distally lock femoral intramedullary nails. Theleft view shows the countersinks and the right view isaligned straight down the axis of the hole.

Fig. 8. Graphical user interface allowing precise C-armalignment. Holes 1 and 2 are represented by small circlesabout the nail axis. The nearly horizontal line represents thenail’s long axis. The larger circle represents the C-arm’sfield of view. In addition, analog and numerical displays(bottom) indicate the angular position of the C-arm relativeto a hole axis for fine-tuning.


phantom covering the range of its intraoperativeuse. This led to a total of 232 visible spheresproviding an equivalent number of error values. Weevaluated all of these visible spheres, includingthose on the border outside the range of the distor-tion correction. The resulting mean error was 0.55mm, with a standard deviation of 0.47 mm and amaximum value of 2.34 mm. The error distributionis given in Figure 11.

Laboratory Pilot StudyDistal locking of femoral nails was chosen to dem-onstrate the feasibility of this technology. A wide

variety of tools and surgical techniques, such aslocked mechanical devices, guidance systems, flu-oroscopic control, and radiolucent drills, have pre-viously been proposed for the accurate and safeplacement of the locking screws. The key problemlies in the significant deformation of the distal endof the nail upon insertion into the femur.22 As aresult, the extensive near-real-time use of the C-arm is commonly accepted. However, besides caus-ing significant radiation exposure to the patient andstaff, the surgical outcome is still not optimal, sincea direct link between the intraoperative images andthe surgical action cannot be provided with currenttechniques.

Using our proposed system, we suggest thefollowing modified surgical procedure: Before in-sertion of the nail, a dynamic reference base withLEDs is attached to the proximal end of the nail.The position of every distal hole is defined with apointing tool, and computer-aided C-arm alignmentensures images are aligned with the axes of thelocking holes (Fig. 9, right). The computer guidesthe C-arm through a 90° rotation and a secondimage is taken on which the counter-sinks of theholes can be easily identified (Fig. 9, left). Real-time image-interactive navigation of a surgical toolis now available to the surgeon: The position of thedrill can be controlled in two images simulta-neously, giving a pseudo–3-D guidance, and thelocking holes can be safely prepared. No additional

Fig. 10. Optoelectronically tracked accuracy verificationphantom. The representations of contained steel spheres inan X-ray image are compared to those determined by thesystem.

Fig. 11. Distribution of the total system error determined in the accuracy study with a total of 232 samples in 40 C-arm shots.


fluoroscopy updates are necessary, resulting in sig-nificantly reduced radiation exposure.

Note that precise computer-guided alignmentof the C-arm with respect to the locking holes isdone under the assumption that deformations ap-plied to the nail during its insertion do not causesignificant rotation to the axes of the locking holesor torsion to the nail, but rather lead to translationaldisplacements of the locking holes, which do notsignificantly affect the C-arm orientation align-ment. This assumption has been shown to be truefor the tibial nail in a preliminary study performedin connection with the development of a guidancedevice.23,24 Furthermore, we have found that smallview misalignments due to nail bending are indi-cated in the C-arm image by a slightly oval appear-ance of the locking hole, and this can be compen-sated for during surgery. It is important to note thatthe image is taken with the nail already bent, sonavigation accuracy—unlike C-arm view align-ment—is unaffected by this phenomenon.

Three test series were carried out: (a) a lab-oratory study on 25 plastic femurs involving a totalof 50 locking holes; (b) anin vitro study using 10human cadaver femurs with 20 locking holes; and(c) an application on two full cadavers with a totalof four locking holes. In all three series, the newsystem allowed easy and successful insertion of thelocking screws. In some cases (20% in series a,10% in series b, and 0% in series c), the drilltouched the nail slightly before entering the hole,with no significant damage or consequences to thelocking procedure. The mean X-ray exposure timeper pair of prepared holes was 1.78 s for series a,1.63 s for series b, and 1.65 s for series c. Thesestudies are described in detail in a forthcomingpaper (Slomczykowski M, Hofstetter R, Sati M,Krettek C, Nolte L-P. A novel computer assistedfluoroscopy system for intra-operative guidance:feasibility study for distal locking of unreamedfemoral nails. Submitted to Journal of OrthopaedicTrauma).

DISCUSSIONA novel kind of computer-assisted freehand navi-gation system has been proposed that is basedsolely on intraoperatively acquired fluoroscopicimages. In addition, rapid acquisition of these im-ages is supported by computer-guided alignment ofthe C-arm. This technology was used to performcomputerized distal locking of femoral nails in alaboratory setup on plastic and cadaver bones. Theresults indicate potential surgical benefits forinvivo use of the system, and the concept of an

entirely precalibrated system keeps its intraopera-tive handling as simple as possible.

In the in vitro studies, the prototype was usedwith caution and a redundant functional check wasperformed before preparation of each surgical ac-tion. Before the concept can be used in clinics on aroutine basis, the mechanical and software compo-nents must be made more sturdy and reliable tofulfil the practical requirements of an intraoperativetool. Hardware and software must also be adjustedfor different C-arm models to prove that themethod yields consistent accuracy.

A key issue during development of the navi-gator was the total system accuracy. Because theC-arm was primarily used as an imaging devicerather than as a means of measurement, geometriclinearity and independence of external influencewere not an issue for most C-arm manufacturers.Thus, the C-arm itself has been found to be themain source of inaccuracy in the concept. Defor-mation of the C-arm frame, as well as image dis-tortions, are reasons for such inaccuracies, andhave to be compensated for to achieve reasonablesystem accuracy. Our compensation methods sig-nificantly decrease—but do not eliminate—the er-rors. With the C-arm model used in this study,remaining errors arising, for example, from vari-able image intensifier distortions caused by exter-nal magnetic fields limited the mean system accu-racy to 0.55 mm and led in rare cases to errors of upto 2.34 mm. We are aware that magnetic distortionscan be a larger source of error for certain C-armmodels; this is the subject of our ongoing work.However, distal locking of the femoral nail (lock-ing hole Ø 5 mm, hole sink Ø 6 mm, drill Ø 4 mm)could successfully be performed with the techniquepresented here. This is because the specially con-cave hole sinks and the conical point of the drillmechanically guide the drill through the hole evenwhen the drill is slightly off target.

A potentially more accurate solution is toupdate the intrinsic and extrinsic calibration in ev-ery C-arm image using marker phantoms within theoperation as proposed in Pietka and Huang,17 andWeese et al.7 However, this has drawbacks in prac-tice, since the optoelectronically tracked, radiolu-cent phantoms have to be placed within the field ofview in the operation. This is cumbersome and, formany orthopedic procedures, impossible: a typicaltradeoff between system accuracy and practical us-ability. Recently, a method for C-arm precalibra-tion was proposed which allows correction for dis-tortions caused by the earth’s magnetic field.25 TheC-arm has to undergo a preoperative “learning”


procedure that determines and stores the distortionsfor all possible C-arm positions relative to theearth’s magnetic field. We recently developed amore practical solution that allows the removal ofdistortions caused by all kinds of magnetic fields;this work will be presented in a future publication.

A second source of error is the position sen-sor that—although the most accurate available—has limited measurement accuracy for rigid bodytransformations. Five consecutive coordinate trans-formations based on the position sensor directly orindirectly affect the accuracy of a tool’s represen-tation in a C-arm image. By increasing the numberof LEDs used and prudently positioning them onthe associated rigid bodies, we were able to in-crease the system accuracy.

The system presented here differs from thatof Viant et al.10 in that the surgeon navigates di-rectly and freehand on multiple computer-alignedC-arm images; the system of Viant et al. uses stereoreconstruction of the locking hole to define a tra-jectory that is carried out with an electromechanicallocking arm.

A fluoroscopy-based system can be seen as acomplement to CT-based systems. Its advantage ofinstant availability without preoperative prepara-tions (CT acquisition, segmentation, and registra-tion) allows it to be used for several applications inthe field of traumatology. It is also open for use inother fields where acquisition of CTs cannot bejustified owing to cost or radiation exposure issues.Since no manual registration procedure is required,fluoroscopy-based navigation may be advantageousin fields where bone structures do not allow recog-nition of the landmarks needed for CT-image reg-istration. Of particular interest is the potential forminimally invasive procedures. Since the C-arm isprecalibrated, a minimally invasive DRB attach-ment is sufficient for keyhole surgical approachesunder multiplanar X-ray control. CT-based meth-ods are more suited in applications where the imagequality of the C-arm is not sufficient or whereattainable perspective views do not provide enough3-D information.

CONCLUSIONA new computer-assisted surgery system has beendeveloped integrating a standard C-arm into a free-hand navigation system. To our knowledge, it is thefirst system allowing surgical navigation basedsolely on multiplanar static C-arm images. Agraphical user interface was developed that allowsreal-time image-interactive guidance of surgical in-struments without further fluoroscopy updates.

Based on the results of the technical accuracy studyand laboratory pilot study presented in this article,two potential advantages of the new computer-based technique, as compared to current manualtechniques, can be identified:

● significantly reduced radiation exposure● improvement of surgical accuracy and safety.

The system holds the promise for a general-ized surgical tool that can be adapted to a widevariety of applications in the field of orthopedicsurgery and offers the possibility of performingminimally invasive approaches.

ACKNOWLEDGMENTS

The authors thank Yvan Bourquin and HeinzWaelti for their help in software programming,Professor C. Krettek for performing the cadaverstudy, and Medivision AG, Oberdorf, Switzerland,for supporting this work by providing electrical andmechanical hardware.

REFERENCES1. Lavallee S, Sautot P, Troccaz J, Cinquin P, Merloz P.

Computer assisted spine surgery: a technique for ac-curate transpedicular screw fixation using CT data anda 3-D optical localizer. J Image Guid Surg 1995;1:65–72.

2. Nolte LP, Visarius H, Arm E, Langlotz F, Schwarzen-bach O, Zamorano L. Computer-aided fixation of spi-nal implants. J Image Guid Surg 1995;2:88–93.

3. Potaminos P, Davies BL, Hibberd RD. A roboticsystem for minimal access surgery. Proc Inst MechEng 1994;208:119–126.

4. Joskowicz L, Taylor RH, Williamson B, Kane R,Kalvin A, Gueziec A, Taubin G, Funda J, Gomory S,Brown L, McCarthy J, Turner R. Computer integratedrevision total hip replacement surgery: preliminaryreport. In: Proceedings of the 2nd MRCAS Sympo-sium, 1995. p 193–202.

5. Hamadeh A, Sautot P, Lavalleé S, Cinquin P. Towardsautomatic registration between CT and X-ray images:cooperation between 3D/2D registration and 2D edgedetection. In: Proceedings of 2nd MRCAS Sympo-sium, 1995. p 39–46.

6. Hamadeh A, Lavalleé S, Cinquin P. Automated 3-di-mensional computed tomographic and fluoroscopicimage registration. Comp Aid Surg 1998;3:11–19.

7. Weese J, Buzug TM, Lorenz C, Fassnacht C. Anapproach to 2D/3D registration of a vertebra in 2DX-ray fluoroscopies with 3D CT Images. In: TroccazJ, Grimson E, Mo¨sges R, editors. Proceedings of the1st Joint Conference on Computer Vision, VirtualReality and Robotics in Medicine and Medical Robot-ics and Computer-Assisted Surgery (CVRMed-MR-


CAS’97), Grenoble, France, March 1997. Berlin:Springer-Verlag, 1997. p 119–128.

8. Brack C, Go¨tte H, Gosse´ F, Moctezuma J, Roth M,Schweikard A. Towards accurate X-ray-camera cali-bration in computer-assisted robotic surgery. In:Lemke HU, Vannier MW, Inamura K, Farman AG ,editors. Proceedings of the International Symposiumon Computer and Communication Systems for ImageGuided Diagnosis and Therapy (CAR’96), Paris, June1996. Amsterdam: Elsevier, 1996. p 721–728.

9. Brandt G, Radermacher K, Lavalleé S, Staudte H-W,Rau G. A medical robot system for orthopedic inter-ventions. In: Lemke HU, Vannier MW, Inamura K,editors. Proceedings of the 11th International Sympo-sium and Exhibition on Computer Assisted Radiologyand Surgery (CAR’97), Berlin, June 1997. Amster-dam: Elsevier, 1997. p 950–955.

10. Viant WJ, Phillips R, Griffiths JG, Ozanian TO,Mohsen AMMA, Cain TJ, Karpinski MRK, ShermanKP. A computer assisted orthopaedic surgical systemfor distal locking of intramedullary nails. Proc InstMech Eng 1997;211:293–299.

11. Hofstetter R, Slomczykowski M, Bourquin I, NolteL-P. Fluoroscopy based surgical navigation: conceptand clinical applications. In: Lemke HU, VannierMW, Inamura K, editors. Proceedings of the 11thInternational Symposium and Exhibition on ComputerAssisted Radiology and Surgery (CAR’97), Berlin,June 1997. Amsterdam: Elsevier, 1997. p 956–960.

12. Tsai RY. An efficient and accurate camera calibrationtechnique for 3D machine vision. In: Proceedings ofIEEE Conference on Computer Vision and PatternRecognition, 1986. p 273–280.

13. Martins HA, Birk JR, Kelley RB. Camera modelsbased on data from two calibration planes. ComputGraphics Image Process 1981;17:173–180.

14. Gembran KD, Thorpe CE, Kanade T. Geometric cam-era calibration using systems of linear equations. In:Proceedings of IEEE Conference on Robotics andAutomation, 1988. p 562–567.

15. Champleboux G, Lavalleé S, Sautot P, Cinquin P.Accurate calibration of cameras and range imagingsensors: the NPBS method. In: Proceedings of IEEE

Conference on Robotics and Automation, 1992. p1552–1557.

16. Boone JM, Seibert JA, Blood W. Analysis andcorrection of imperfections in the image intensifier-TV-digitizer imaging chain. Med Phys 1991;18:236 –242.

17. Pietka E, Huang HK. Correction of aberration in im-age intensifier systems. Comput Med Imaging Graph-ics 1992;16:253–258.

18. Rudin S, Bednarek DR, Wong R. Accurate character-ization of image intensifier distortion. Med Phys 1991;18:1145–1151.

19. Mansbach P. Calibration of a camera and light sourceby fitting to a physical model. Comput Vis GraphicsImage Process 1986;35:200–219.

20. Northern Digital, Inc. Optotrak 3020, the 3D motionmeasurement system. System Specifications, April1994.

21. Schauer D. Entwicklung von Instrumenten und Ap-paraten fu¨r die computerassistierte orthopa¨discheChirurgie. Diploma thesis, Institute for Electronics,TU Berlin, 1997.

22. Knudsen C, Grobler G, Close R. Inserting the distalscrews in a locked femoral nail. J Bone Joint Surg1991;73-B:600–601.

23. Krettek C, Mannß J, Ko¨nemann B, Miclau T, Schan-delmaier P, Tscherne H. The deformation of smalldiameter solid tibial nails with unreamed intramedul-lary insertion. J Biomech 1997;30:391-–394.

24. Krettek C, Ko¨nemann B, Miclau T, Ko¨lbli R, Mach-reich T, Kromm A, Tscherne H. A new mechanicalaiming device for the placement of distal interlockingscrews in femoral nails. Arch Orthop Trauma Surg1998;117:147–152.

25. Brack C, Burgkart R, Czopf A, Go¨tte H, Roth M,Radig B, Schweikard A. Accurate X-ray-based navi-gation in computer-assisted orthopaedic surgery. In:Lemke HU, Vannier MW, Inamura K, Farman AG,editors. Proceedings of the 12th International Sympo-sium and Exhibition on Computer Assisted Radiologyand Surgery (CAR’98), Tokyo, June 1998. Amster-dam: Elsevier, 1998. p 716–722.


Documents

Fluoroscopy as an imaging means for computer-assisted surgical navigation